Transcript Use of probabilistic climate change information for impact

Use of probabilistic climate change
information for impact & adaptation work
(ENSEMBLES & UKCP09)
Clare Goodess
Climatic Research Unit
[email protected]
WCRP workshop on regional climate change
ENSEMBLES regional scenarios portal
www.cru.uea.ac.uk/projects/ensembles/ScenariosPortal
Some examples of PDFs
All these PDFs are conditional (on A1B, underlying assumptions….)
And subjective – do not sample the ‘full’ uncertainty range.............
Some ENSEMBLES conclusions:
• ‘Subjective’ not ‘objective’ probabilities
• Users need to know underlying inputs,
assumptions & methods
• Underlying assumptions tend to be harder to
explain & understand than for deterministic
scenarios
• A wide PDF does not mean that all uncertainties
have been extensively or uniformly sampled
• Good communication & user guidance are
essential
UKCP09 national climate change projections
probabilistic
projections
over land
weather
monthly/
generator
daily
seasonal
time
PDFs
series
http://ukclimateprojections.defra.gov.uk/
PDFs summer Tmax, East of England, 2020s & 2050s
CDFs summer Prec, East of England, 2020s & 2050s
Users preferred numbers to graphs – but not too many.........
Summary of UKCP09 projected changes for the East of England administrative region. 10%, 50% and 90%
probability estimates are given for the medium emissions scenario. The Wider Range indicates the 10% to
90% range across the low to high emissions scenario (each of which are considered as equally likely as the
medium emissions scenario). Temperature changes in °C, precipitation changes in %.
UKCP09 key messages:
Under medium emissions for the 2050s, the central estimate of increase in summer mean
daily maximum temperature is 3.4°C; it is very unlikely to be less than 1.3°C & is very
unlikely to be more than 6.0°C. A wider range of uncertainty is from 1.1°C to 6.8°C.
Under medium emissions for the 2050s, the central estimate of change in summer mean
precipitation is -17%; it is very unlikely to be less than -38% & is very unlikely to be more
than +6%. A wider range of uncertainty is from -40% to +14%.
•
Zero or very small changes in annual precipitation (rainfall and snow) mask contrasting
seasonal changes: with wetter winters & drier summers. In spring, the most likely change
(i.e., the 50% probability – or the mode/peak of the PDF) is ‘no change’, i.e., zero, with
some indication of a rather small increase in autumn.
•
The central estimate for the medium emissions scenario is -7% in summer, with the
change likely to be between a 24% reduction & a 15% increase for the 2020s. For the
2050s, the central estimate for the medium emissions scenario is -17% in summer. The
wider range indicates that the summer decrease is very unlikely to be more than -40%,
although the upper range is positive (+14%).
•
Another, and perhaps simpler, way of expressing the summer changes is that, for the
2020s there is a 60 to 70% probability that total summer rainfall will decrease, rising to 74
to 85% probability for the 2050s.
Indices based on UKCP09 weather generator output
(100 x 30 years – sampled change factors from PDFs)
Changes in counts of GDD (growing degree days) for Lowestoft.
The left plot shows the increase in mean annual values (with upper and lower 95% confidence
intervals) for the 2020s (medium emissions) and 2050s (low, medium and high emissions).
The right plot shows the seasonal increase in mean values for the 2050s (medium emissions – red
boxes) relative to the baseline period (purple boxes). The variation in counts is shown by the upper
and lower edges of the boxes (1 standard deviation above and below the mean) and the T ‘whiskers’
(2 standard deviations above and below the mean). Wider maximum and minimum values are
represented by black crosses.
Design Summer Years (DSYs) and
Test Reference Years (TRYs) for use
in building performance models
DSY: April-Sept mean T
Middle of the upper quartile
TRY: The most ‘normal’ year
selected for each month
Challenges for the community:
• Current DSYs/TRYs don’t represent the present
CIBSE TM36, 2005
• Need to handle climate variability as well as change
• Need to work with uncertainty (e.g.,multiple emissions scenarios)
• And now need to work with probabilities (UKCP02 to UKCP09)
• Range of ‘users’: engineers/consultants, architects & their clients
DSYs from UKCP09 weather generator output
100 x 30 years
London
Medium emissions
Method 1: Calculate the metric for each 30
year period & then take 50th percentile of 100
values
Method 2: Calculate 50th percentile from 3000
metric values
Dashed lines: CIBSE TM48
Solid lines: UKCP09 wgen
Thanks to Vic Hanby & Stefan Thor Smith
EPSRC PROCLIMATION project, DMU
There are benefits in using
probabilistic projections, but......
WCRP workshop on regional climate change